Newton-Raphson algorithm for 3-phase load flow

R. G. Wasley, B.Sc. Tech., M.Sc, Ph.D., and M. A. Shlash, M.Sc. Indexing terms: Electrical engineering application of computers, Load (electric), Transmission-network calculations ABSTRACT One of the problems encountered in power-systems operation is the generation of unbalanced voltages and currents in the presence of long transmission lines with few or no transpositions, including possible unbalances arising in source and load conditions, or indeed any items of plant such as shunt and series reactors. To improve or investigate these unbalance effects in any detail, a 3-phase-load-flow solution that allows representation of all possible unbalances as they exist in the power-systems network without making any assumptions is essential. Consequently, a working Newton-Raphson algorithm is presented, which forms the basis of a computer program developed for the specific purpose of solving the 3-phase-load-flow problem. An indication of program facilities is given, together with sample results for illustration. LIST OF PRINCIPAL SYMBOLS Vk = Ek + j F k = array of phase voltages at node k Ik = CRk + jCIk = array of phase currents at node k NCk = NCRjt + NCIk = array of nodal currents at generator node k Sk = P k + jOjf = array of single-phase powers at load node k = total 3-phase power at generator node k SLk = PLk + jQLk = array of local loads per phase at generator node k Y = G + jB = network admittance matrix EMk = array of machine e.m.f.s at generator node k ILk = array of local load currents at generator node k ZGk = machine impedance matrix at generator node k YGk = admittance matrix that corresponds to the inverse
of ZG k

time. In this respect, the Newton-Raphson method possesses good convergence and reliability properties, as well as great generality and flexibility, which makes it very convenient when representing various conditions and constraints in the system. 4 ? 5 Like any other matrix method, however, it requires excessive storage and computation unless efficient sparsity techniques are employed. The method is also very sensitive to good starting conditions. The use of a suitable starting process can reduce computer time substantially, as well as ensure convergence.6 Based on these considerations, therefore, a general method of applying the Newton-Raphson algorithm is described for the purpose of solving the 3-phase-nonlinear-load-flow problem. For illustration, sample results are also presented, as obtained by a specially developed computer program, which incorporates sparsity techniques and similar facilities found in programs designed for the single-phase solution. Some features of the 3-phase program are as follows: (a) loads are entered as constant power sinks in a specified balanced or unbalanced state (b) lines in the network are taken in their natural unbalanced mutually coupled form (c) any line discontinuities can be included; e.g. transposition, series and shunt reactive compensation, phasing arrangements at busbar ends for both singleand double-circuit lines, as well as a particular line operation, like one circuit of a double-circuit line open at one end (d) full machine representation is permitted, so that means are available for including saliency in synchronous machines and also induction-motor loads (e) allowances are made for the investigation of different methods of voltage control at generator busbars (f) facilities for online tap changing of transformers are available, as well as provisions for including reactive power limits at the generator busbars

INTRODUCTION

As the number of extra-high-voltage lines increases in transmission networks, particularly in areas where large hydroelectric stations supply large and remote load centres through lines with few transpositions, the unbalance effects of these lines with any possible unbalance in load or source must be properly analysed. The unbalance effect may not be especially significant as far as the network itself is concerned, but in terms of the individual components of the network it can be extremely important. For instance, the presence of negative-sequence current at the generator terminals will give rise to heating in its rotor. Furthermore, because of unpredictable current distributions, protective relays could malfunction and active power loss could be comparatively higher than expected because of parallel untransposed transmission lines. 1 For day to day operation, the usual load-flow solution is based on perfectly balanced 3-phase networks operated under balanced 3-phase generation and load conditions; i.e. a single positive-sequence network is involved. In its more general form, the sequence component frame has also been applied in an attempt to include unbalance effects. 2 ' 3 This approach is restrictive, however, and was applied partly because of the difficulty of representing transformers, generators and induction motors in the phase frame of refference, and partly because of the limitations of the method adopted at the time. In the present work, development of the 3-phase-load-flow algorithm is associated with the phase frame of reference entirely, as this permits a full and comprehensive treatment, although the problem of fully representing machines, lines and transformers in the phase frame of reference for incorporation in the load-flow program will be left to a future paper. The choice of method for any load-flow solution depends on many factors, the most important being convergence characteristics, storage requirements, reliability and computer

(g) as all branch elements are represented in general 2port-admittance-matrix form, it is possible to incorporate any method of neutral-point earthing and connection of transformers. Transformer unbalances can also be included.

PROBLEM FORMULATION

The main object of the load-flow solution is to obtain the individual phase voltages at all busbars in the network corresponding to specified system conditions. As voltage magnitudes and angles are involved, it follows that, for each busbar, six independent constraints are required to solve for the six unknowns. These constraints are different for different types of busbar; the two main types of which are load and generator busbars, with the slack busbar being treated as a special case of the generator busbar. 2.1 Load busbars The active and reactive powers for each of the three phases are specified separately, so that balanced or unbalanced conditions can be considered as required. Thus, in a straightforward manner, these will provide the necessary six equaPROC.IEE, Vol. 121, No. 7, JULY 1974

Paper 7178 P, first received 27th December 1973 and in revised form 14th March 1974 Dr. Wasley and Mr. Shlash are with the University of Manchester Institute of Science & Technology, PO Box 88, Manchester M60 1QD, England
630

tions for solving corresponding voltage magnitudes and angles. Generator busbars The total 3-phase active power output is specified together with one magnitude. The latter may correspond to one of the three phases, or to any combination of the 3-phase voltages, e.g. the positive-sequence voltage, depending on the voltage-regulation scheme that is adopted in a particular generator or group of generators. These specified quantities will only give rise to two constraints, however, and so the remaining four constraints are obtained by making use of the three voltages behind the machine impedance, which are balanced in both their magnitudes and angles for a balanced design of the generator windings. This is true, as the excitation is in this case equally distributed to the three windings of the machine. For the slack busbar, the active power output is not specified, but obtained in solution. This is because the active power loss in the network is initially unknown. However, the angle of the voltage controlled by the corresponding generator can be specified and used as a reference. MISMATCH EQUATIONS 3 As in the single-phase case 7 the 3-phase-load-flow problem can be solved by the Newton-Raphson method using a set of nonlinear equations to express the various specified constraints. Thus, by using the constraints for those types of busbar mentioned above, suitable mismatch equations are derived to provide a basis for defining the elements of the 3-phase Jacobian matrix. Load busbars The per-unit complex power associated with phase p at load busbar k, as specified, can be expressed in terms of busbar voltages at all nodes n to give the equation -S k P=V k P(I k P)*=V k P I I (Y ki Pm Vi ^)*
i = im=i

Pk = Re I Vkm(Ikm)* + I
m=i m=i

PLkm

2.2

where Re denotes the 'real part of. Again taking the real and imaginary parts of the complex elements, = Re (E k m + j
m=i

- jCIk) +

PL k
m=i

leading to the mismatch equation = Pk Z ( E k m CR k m +

m=i

(3)

for P k the specified active power, (b) Terminal voltage mismatch This mismatch equation depends on the voltage regulation scheme; but supposing that the automatic voltage regulator of the generator, in the first instance, is controlling one phase only, e.g. the magnitude of the first phase, it follows that the corresponding mismatch equation is given by
A | V k i | = |V k l|
1

(Fkl)2}V2

(4)

where | Vj^ 1 is the specified magnitude. An alternative equation controlling the positive-sequence voltage is introduced later. (c) Internal voltage mismatch Since the generator voltages are taken as balanced, they must maintain a positive sequence distribution as defined by the equations aEMk2 = EM k ! = (a)2EMk3 where a is the usual 120 complex operator. These equations, however, must now be considered in terms of machine representation. They may be used in conjunction with detailed machine models including the effects of saliency and other machine unbalances, as will be shown in a future paper, but, for the present, it is sufficient to consider a simplified model leading to the equation EMk = Vk + ZGk(Ik + ILk) where IL k represents local load current, and ZG^ is a 3-phase-generator impedance matrix derived from sequence impedances8 as shown in Appendix 12. On the basis of this simplification, it now proves expedient to consider the latter equation in terms of an equivalent current source to define the following: NCk = YGkEMk =
IL k

3.1

where Y represents the network admittance matrix, and the negative sign appears because the nodal current entering the network is taken as positive. Alternatively,taking the real and imaginary parts of the complex elements,

i=i

1t

m=i

so that separation leads to the mismatch equations + EkP

i=i m=i

- B ki Pm Fi m)

(1)
i=i m=i

where YGj^ is the admittance matrix corresponding to the inverse of ZGk. Alternatively, we may write for each phase p at generator busbar k NCkP = Y ki Pm Vi m + ILkP
i=i m=i

+ FkP
i=i m=i

" EkP Z Z
i= l m=i

(2)

where p takes the values 1, 2 and 3 corresponding to the three phases involved.
3.2

where the additional machine admittance matrix is now understood to be absorbed in the diagonal block of the network admittance matrix. Because of the special form taken by YGk (Appendix 12), the balanced form of EMk must also be reflected in NCk. Consequently, from aNCk2 = NCk* = (a)2NCk3 the voltage constraints are applied, i.e. for the simplified model, to provide current mismatch equations. Separating into real and imaginary parts these are defined by the following equations: ANClk = - V2NCRk2 - V3/2 NClk2 ANC2k = V3/2 NCRk
2

Generator busbars For the generator busbars, there are three kinds of mismatch equations as follows:

(a) Power mismatch The total active power at generator.busbar k including local load is given by the equation PROC.IEE, Vol. 121, No. 7, JULY 1974

(5)

(6) 631

ANC3k = - V2NCRk3 + V3/2 NCIk3 ANC4k = - V3/2 NCRk3 - ^NCl k 3 -

(7) (8)

AE k 2 H
n

AF k 2 N12 -H12
N22

AE k 3
H

AF k 3 Ni3
~H13

N11

12

13

Nn

-Hn
N21

Ni2
H

Ni3
H

4 JACOBIAN MATRIX The 3-phase Jacobian matrix comprises matrix blocks of the order 6 x 6, so that, in the following, the derivations of diagonal and off-diagonal blocks for load, generator and slack busbars are given using the defined mismatch equations. On this basis, from starting voltages, either assumed . or obtained by some approximate method, mismatch quantities AM and their derivatives forming the Jacobian matrix J are used to find incremental voltage quantities AV. Thus, by solving the linear set of equations expressed by the matrix equation AM = - JAV the same iterative procedure followed in single-phase-loadflow solutions7 is implemented to reach a satisfactory state of convergence. Voltages are updated using the equation V = V + AV For a system of n busbars, 3-phase analysis introduces a Jacobian matrix of the order 6n x 6n, which means that, even in comparatively small networks, the matrix is sufficiently large to suggest a need for careful programming. Therefore, in taking steps to reduce costs of computation, the authors' program also incorporates the concepts of ordered elimination and bifactorisation.9 These steps are not considered further, but essentially the 6 x 6 blocks of the Jacobian matrix are processed in the same manner rvs the scalar elements of the Jacobian matrix used in single-phase analysis.
4.1

21

22

23

N23
~H23
N33

AQk^

N21
H

~ H 21
N31

N22
H

"~H22

N23
H

31

32

N32
~H3 2

33

N31

-H3i

N32

N33

~H3 3

(b) Off-diagonal block Taking load busbar k, eqns. 1 and 2 are differentiated in turn with respect to Ei m and Fi m , for all m and p from 1 to 3, and for all busbars connected to k, i.e. for all corresponding values of i, except i = k. Thus 3APkP =H
pm

- E k PB ki P m = N p m

9 E .m

The off-diagonal blocks associated with each load busbar i 5 * k will therefore take the following form: AEjl
H'n
Nil

Load busbars The corresponding diagonal and off-diagonal blocks of the Jacobian matrix are both derived from the mismatch equations (eqns. 1 and 2) as follows: (a) Diagonal block Taking load busbar k, eqns. 1 andm2 are differentiated in turn with respect to E k m and F k , i.e. the real and imaginary parts of voltage V k m , for all m and p from 1 to 3. Thus 3APkP = EkpGkkPP + F k p Bkk PP + = H pp for m = p
APk

AFi1
N'n
-H' 11
N2i

AE^ 2
H

AF;2
N'12
~ H 'l2
N22

AE;3
H

AFJ 3 N' 1 3 "" H l3

'l2

'l3

N' 1 2
H

N'13
H

21

22

23

N23
~H2 3
N33

N21
H

~H2 1
N31

N' 2 2
H32

~ H 22

N23
H

31

N'32
~H32.

33

AQk 4. 2

N'31

~H31

N32

N33

~H33

+ FkPBkkPm = H p m for m * p -a APkP

Generator busbars

= Fk^fck^ - EifPBi&PP + CJ. p =

for m = p PP

The first two rows of both diagonal and off-diagonal blocks are derived from the power and voltage mismatch equations, eqns. 3 and 4, respectively, whereas the remaining rows are derived from the current mismatch equations, eqns. 5, 6, 7 and 8, respectively. (a) Diagonal block Taking generator busbar k, eqn. 3 is differentiated with respect to E k m and F k m , for all m = 1 3, where

= Npm for m ^ p

- CIkP = N pp for m = p - EkPBkPm = N p m for m * p

CRkm = I I

iq - B ki mq Fi q)

i = i q=i

CIkm = Z CRkP = - H,PP for m = p - FkPBkkPm = - H p m for m * p The diagonal block associated with each load busbar will therefore take the following form:
632

i = i q=i

I (Gki^ +

to give
3

m=

"q?i
3

- CR k m = PE m - CI k m = P F m PROC, IEE, Vol. 121, No. 7, JULY 1974

q=i

Note that both the local active power and P^ are constants leading to zero derivatives. Similarly, for single-phase voltage control, eqn.4 is differentiated to give the corresponding elements of the Jacobian matrix as = - E k i{ (Ek 1 ) 2 + (P k i)2}-i/2 = 0 for m = VF for m = 1
for and

V3

LC1 1 = SE 1 LC2 x =

3Ek

VE for m = 1 LC3 1 = SE 1 LC4 1 = S F 1

1 Ji LC12=|-SE2 + ^-SF2 J 1 Ji LC2 2 = jSF2 - ^ S E 2 LC3 2 = 0 LC4 2 = 0

LC1 3 = 0 LC2 3 = 0 ^-!

LC3 3 = j SE 3 -

LC4 3 = j SF 3 + ^ - J

m = 0 for m

= [{(E k m ) 2 m

Before differentiating eqns. 5-8 with respect to E k and F k m , for all m = 1 3, it is noted that local load current is defined so that IL k P = (SL k P/V k P)* f o r P = 1, 2, 3 Therefore, it can be shown that the nodal currents in real and imaginary forms are given by the equations

{(E k ) 2 +(F k ) 2 } 2 = {[(Fk)2 [{(Ekm)2 From the derivatives given, it now follows that the diagonal block of the Jacobian matrix corresponding to a generator node will take the form

APk A|V k i| ANCl k ANC2k ANC3k ANC4k

E 1

F l

PE2 0
+

PF2 0

PE3 0

PF3 0 C 2 3 + LC2 3 Cl3 -LC13

VE Cl. + LCl,

VF

C2 1 cii-

LC2X LC1 X

C1 2 + LC1 2 - C 2 2 + LC2 2 C3 2 + LC3 2 - C 4 2 + LC4 2

C2 2 + LC2 2

C1 3 + LC1 3 - C 2 3 + LC23 C3 3 + LC3 3 - C 4 3 + LC43

c i 2 - LC1 2
C4 2 + LC4 2 C 3 2 - LC3 2

w0

" t *

JJVOI

JL

C4 1

LC4X

C4 3 + LC4 3
C33 - L C 3 3

C 3 , - LC3 X

NCR k P= f

Z (GkiPmEim - B k i Pm F i

(b) Off-diagonal block In this case, eqn. 3 is differentiated with respect to all E^ and F i m , but not i = k, to give for all m = 1 3 =E(E k ( iG k i q^ + q=i

(PL k PE k P + QL k PF k P)/{(E k P) 2 + (F k P) 2 }
n 3

= PEo

NCI k P=
q=i

+ (PL k PF k P - QL k P)/{(E k P) 2 + (F k P) 2 } Therefore, from eqns. 5-8 it follows that, for m = 1 3, 9ANCl k = C lm m + LClm 3Ekm 3ANC2k 3ANCl k 3F k m 3ANC2k + LC2 m
m

Similarly, taking the mismatch equation, eqn. 4, it follows that, again for single-phase voltage control,
= 0

a A I Vk 1

m m

= 0

On the other hand, eqns. 5-8 lead to the derivatives = Clm - LClm 3 ANClk 3 ANClk

C2

3ANC3k

= C 3m m + LC3 LC3 m 9 ANC4k

3ANC2k 3Eim
=C3m-LC3m

3ANC2k

3 ANC4k

3ANC3k 3Ei
m

3ANC3k aFim 3ANC4k

where
2m

3ANC4k
lm

where
G 2m_ i l 2 ki 2

PROC.IEE, Vol. 121, No. 7, JULY 1974

633

2m JL ft. + 2 B ki. 2m _ 1?_ 2 Gi ki +

.im

In this way, the voltage magnitude mismatch equation becomes A|X k | = |X k | ~{(XX)2 + (X 2 ) 2 }i/2 (9)

iKi3m
i

.3m

V3_ 2

a-

3rn

for |Xkl , the specified voltage magnitude. Eqn. 9 is now used to derive those elements of the Jacobian matrix previously found from eqn. 4. Clearly, the derivatives of A|Xkl with respect to all E i m and F j m (i * k) are zero, whereas, with respect to E k m and F k m , m = 1 3, we have Xx 3|X| 3A|X k | 6|X| 3A|Xk| X1 + V3X2 6|X| for (X 2 ) 2 }V 2 3 AI X k I 3A|X k | 6|X| -V3X X + X 2 6|X| 3A|X k l X2 3|X|

Consequently, for i * k, off-diagonal blocks of the Jacobian matrix corresponding to a generator node will take the form AFil
AP k

AEi 2 PEO2 0
Cl2

AFi 2 PFO 2
0

AE^

AF t 3 PFO3 0 C2 3
Cl3

PEOi 0

PFOi 0

PEO3
0 Cl3

A|V k l| ANClk ANC2k ANC3k ANC4k 4.3

cii
no'

C21
Cl'i

C2' 2
Cl2

-C23 C3 3 -C43

C3;

C4i C3i

C3' 2 -C4'2

C4 2 C3 2

C4 3
C3' 2

-C4i

Slack busbars These equations can also be used in the case of positivesequence control at the slack busbar, so that the constraint on |X k | replaces the constraint A E ^ = 0. Otherwise everything remains the same, as even though the magnitude of phase one voltage is no longer specified, the constraint AFj^1 = 0 may still be applied provided the angle of phase one voltage is set to zero, corresponding to Fj^ 1 = 0. If, however, the positive-sequence voltage angle is used as a reference, in the case of the slack busbar, the constraint on Fj^ 1 may be replaced by AW = X o - X1 tan (10)

As already mentioned slack busbars are treated as special generators. Thus, instead of the active power requirements, it is the angle of the controlled voltage that is specified because it is taken as a reference. The controlled voltage is also specified, of course, so that for phase one control at the slack busbar k, it follows that
= 0

This means that, since Ek 1 and Fk 1 require no formal solution, there is a corresponding redundancy of equations leading to an appropriate reduction in the order of the Jacobian matrix. However, from a programming point of view, apart from the fact that there is little to be gained by taking advantage of the reduction, it proves much more convenient if all blocks of the Jacobian matrix are kept in the general 6 x 6 form. In effect, the reduced Jacobian matrix is expanded to accommodate the trivial relationship AEk
1

where 8 is the specified angle. Thus, derivatives of AW with respect to all E i m and F j m (i * k) are zero, and derivatives with respect to E k m and F k m , m = 1 3, are given by

a AW a AW

tan e

a AW

1 0

0 1

AEk

V3 + tan 0
6

AF k l

AFk 1

a AW
3Fk2

1 + V3 tan 6 6 1 V3 tan 9

Therefore, when considering the slack busbar k, the normal generator mismatch elements APk and AlVk 1 ! are effectively replaced by AEk1 anc^ ^Fj^ 1 , respectively, and set to zero, as are all row and column elements of the Jacobian matrix associated with the AEk1 anc * AFjj 1 of the slack busbar, except the two diagonal elements. These are set to unity and so satisfy the above equationalthough the choice is in fact arbitrary provided indeterminate matrix inversion is avoided.

a AW

- V 3 + tan e

a AW

For simplicity 6 may be set to zero. Finally, it should be noted that, in lifting the constraints AEk1 = 0 and AFk1 = 0 at the slack busbar, the derivatives of all other mismatch elements with respect to the slack Ek 1 and Fk 1 must in the usual way take the general forms given in Sections 4.1 and 4. 2.

ALTERNATIVE VOLTAGE CONTROL

6 REACTIVE POWER For a solution that is acceptable in practical terms, the reactive power limits of the generator units should not be exceeded. Therefore, as in single-phase studies, any generator node, at which the reactive power limits are violated during the course of solution, is considered from that point on as a node having no voltage magnitude constraint^ but instead a reactive power constraint equal to the violated limit. In the 3-phase case, the total reactive power is required as given by

So far the 3-phase Newton-Raphson algorithm has been derived on the basis of single-phase voltage control at generator nodes, whereas other schemes, e.g. positive-sequence voltage control, are also of interest. Thus, as an alternative scheme, consider the positivesequence voltage at generator node k to be defined, so that X k = VafVk1 + aV k 2 + (a) 2 V k 3 ] to give the magnitude IX k l = { ( X 1 ) 2 + ( X 2 ) 2 } l / 2 where the real and imaginary parts of Xk are, respectively, E k 3 ) + ^-(F k 3 X 2 --
634

Qk = Im I Vkm(lkm)* + I QL k m
m= i m= 1 where Im denotes the 'imaginary part of and QL represents local reactive power. Taking the real and imaginary parts PROC. IEE, Vol. 121, No. 7, JULY 1974

F k 3 ) + f-(Ek 2 - E k 3 )}

of the complex elements,

3 3
jCIkm)

Qk = Im I (E k m + jF k m)(CR k -

QL k m

m=i which leads to the mismatch equation

= Qk -

m=i

Transformers are represented on the basis of leakage reactance values,and the synchronous machines by balanced impedance matrices obtained from the sequence impedances in the manner described in Appendix 12 (Section 3.2). In the case of transformers Tx and T 4 , it should be mentioned that tap settings of 4*6 and 1-5%, respectively, are also involved. Shunt reactors are provided and it is noted that system data are presented in Table 1.
circuit 1

m=i

(Fk m CR k m - E k m C I k m + QL k m)

(11)

for Qif, the specified reactive power limit. It now follows that, noting that the local reactive power as well as Qk is constant, corresponding elements of the Jacobian matrix are established by differentiating eqn. 11 to give: (a) Diagonal block
3E k m CI k
q=i generator X A-Y Y-Y

Y-Y

load

- CR k m
fo r m = 1 3 .

Fig.l System studied TABLE 1 SYSTEM DATA (REACTANCE VALUES ON 1000 MVA BASE) Generator sequence reactances z = 0*7500, z 2 = 0-1725, z 0 = 0-0900 p.u. Synchronous-compensator sequence reactances zx = 1-2250, z 2 = 0*2800, z 0 = 0-1000 p.u. Transformer leakage reactances T x : 0-1710, T 2 : 0-0700, T 3 : 0-0837, T 4 : 0-1167 p.u. Shunt reactors X1: 320 MVA, X2 : 600 MVA Horizontal line configuration Mean height of phase wires = 16-0 m Mean height of earth wires = 30*0 m Separation between phase wires = 14-3 m Separation between earth wires = 13-8 m Separation between circuits = 60*0 m (centre to centre) Phase bundle : quad 23-5 mm Dove a.s.c.r. (spacing 45 cm) Earth wire: 12-5 mm steel Line section lengths 2-3 : 304, 3-4: 420, 4-5 : 314 km Earth resistivity = 100 Sim System frequency = 50 Hz As a further point of interest it is noted that the practical situation demands A Y connections for the transformers T x and T 4 . Hence, in formulating equivalent 2-port admittance matrices, a modification to the basic Y Y representation is simply obtained by introducing an appropriate connection matrix that ensures that the phase currents on the generator side of the transformer add to zero. Suitable scaling by the factor V3~is also required when working in per-unit values, but such details are outside the scope of the present paper and it suffices to say that, once equivalent 2-port matrix equations are established for any type of transformer branch and incorporated into the complete network admittance matrix, the fundamental algorithm is applied as described in the paper without any modification whatsoever. Generator busbar 1 is treated as a slack busbar having a controlled R phase voltage magnitude of 1*0 p.u. Similarly, the R phase of the synchronous compensator (P = 0) is also maintained at 1-0 p.u. Thus, for a balanced distribution of the total 3-phase load amounting to 763*8 MW and 2526 MVAr at busbar 6, branch flows resulting from a load-flow solution are presented in Table 2. Also given under 'In' are the corresponding ratios of negative-to-positive sequence currents expressed as a percentage.
635

(b) Off-diagonal block

im

=- t
q=i

for m = 1 3 and i = 1, 2,.. ., n, but

7 STARTING PROCESS

In effect, the described algorithm considers each phase at any one busbar as a separate node, being linked to all such nodes through relationships that correspond to the given network. Hence, in practice, the individual voltages at a busbar can be unbalanced with respect to each other both in magnitude and angle. For the usual steady-state conditions, however, it can be expected that these voltages will have a distribution that, for the most part, will be of positive-sequence form. Consequently, a single-phase starting process can be applied, so that, having obtained the single-phase estimates, each of the three corresponding phases takes the estimated voltage, but an angle that is shifted 0,120 or 240 from the estimated angle depending on the phasing arrangements. It should be noted that, in the case of A Y transformers in the network, it also helps a great deal to adjust angles on the high-voltage side by 30 or 30 according to the connections being used. If starting values appear to be a problem, e.g. in highly unbalanced situations, a 3-phase starting process could be adopted by assuming active power as equally distributed in the three phases of the generator nodes to permit a 3-phase d.c. load-flow solution. In this way, the voltage angles for each phase at all busbars could be estimated taking into account line unbalance.
8 SAMPLE RESULTS

For the purposes of demonstration, the derived NewtonRaphson algorithm is implemented in a 3-phase-load-flow program to obtain sample results corresponding to the 500 kV network of the double-circuit configuration (Fig. 1). Because of the long lengths, each homogeneous section of line is represented in 2-port admittance matrix form using modal theory 10 and connected in accordance with the defined phasing arrangements involving both midpoint transpositions and busbar phase rotations.
PROC. IEE, Vol. 121, No. 7, JULY 1974

For the same conditions, busbar voltages are presented in Table 3, where the usual 120 and 240 phase shifts of the Y and B phases are suppressed as a matter of convenience. Because of the AY connections of the machine transformers, it is seen that angles on the high-voltage sides of these units are shifted by approximately 30. This Table also includes a measure of voltage unbalance, so that the maximum voltage difference between phases is expressed as a percentage of the corresponding mean voltage. As a case for comparison, branch flows and busbar voltages are also presented in Tables 4 and 5, respectively, but for a situation in which the second circuit of line sections 3-4 (Fig. 1) is disconnected from busbar 4. Here, a worse case of unbalance arises in redistributed form because of the additional line unbalance. Furthermore, the rather poor voltage profile along the system is noted, leading to a voltage reduction of about 8% at the load bus compared to a previous value of about 3% . The basic reason for this, of course, is associated with the increase in equivalent reactance between busbars 3 and 4.
TABLE 2

The possibility of an apparent gain in active power flow in any one phase is indicated. Hence, by way of example, Table 4 shows an increase in the active power flow in phase 2; i.e. between buses 1 and 2 it increases from 277*22 to 280*78 MW. This behaviour is inevitable, of course, in an unbalanced situation, although it is noted there is always a net loss in total active power unless, as in this case, no losses are included. In effect, the transformer between busbars 1 and 2 is represented in terms of reactance only, so that here the total active power flow remains unchanged. As another point of interest, it is seen that, at the synchronous compensator busbar, individual phases may carry active power although total 3-phase power satisfies the imposed constraint P = 0. Finally it is worth mentioning that total network losses in the first study amount to 37-94 MW compared to 54*35 MW in the second study, which is to be expected because of the net increase in system unbalance.

BRANCH FLOWS AND PERCENTAGES OF NEGATIVE TO POSITIVE SEQUENCE CURRENTS FOR NORMAL LINE-IN CONDITIONS

Sending end Sending busbar Receiving busbar Phases Active

MW 1

Receiving end
In %

Reactive
MVA

Active
MW

Reactive
MVA

In %

R Y B R Y B R Y B R Y B R Y B R Y B R Y B

267-09 268-98 265-67 26551 268-97 267-27 261-58 261-16 267-67 261-95 25599 256-93 253-90 256-52 253-37 -0-70 1-92 1-23 -1-67 0-09 1-59

-68-48 -66-19 -65-00 -105-25 -107-72 -103-66 -129-90 -140-78 -138-41 -96-58 -95-14 -104-36 69*58 7324 71-42 -29-79 -26-38 -27-83 29-32 26*70 29-46

0-71

265-51 268-97 267-27 261*58 26116 267*67 261*95 255*99 256*93 253-90 256-52 253-37 253-90 256-52 253-37 -0*70 1-92 -1-23 0-70 -1-92 1*23

-105-25 -107*72 -103-66 66-40 56-81 58-91 98-73 100-76 90-97 69-58 73-24 71-42 54-41 57*82 56*37 -30-03 -26-56 -28-04 30-03 26-56 28*04

0*71

0-71

1*41

1-49

1-38

1-28

0-64

0-64

0-64

6-45 6-45

6-45

6-45

TABLE 3 BUSBAR VOLTAGES AND PERCENTAGES OF VOLTAGE UNBALANCE FOR NORMAL LINE-IN CONDITIONS R phase Y phase Angle
deg

B phase Angle
deg

Busbar
1 2 3

Magnitude p.u. 1-000 0-995 0*991 0*988 0-979

0*966 0*974

Magnitude p.u. 0-999 0-996 0-994 0-990 0-985 0-970 0-977 1-002

Magnitude p.u. 0*998 0*993 0*993 0*988 0*983 0*969 0-977 1-000

Angle
deg 003

V %

0*00 -37*47 -4511

-56*25

-0*07 -37*65 -45*84 -56*17 -64*43 -67*66 -67*69 -37*81

0*16 0*32 0*33 0*15 0*53 0*46 0-37 0-24

-37*61 -45*92 -57*02 -64*72 -67*92

-67*90

4 5
6

-64-59 -67-82 -67-81 -37-72

7
8

1*000

-37*88

636

PROC. IEE, Vol. 121,No. 7, JULY 1974

CONCLUSIONS The development of a Newton-Raphson algorithm is described which permits the analysis of unbalanced powersystem networks in the phase frame of reference. As such, the algorithm forms the basis of a computer program which can be used to establish 3-phase load-flow solutions.
9

with different methods of voltage control, tap staggering and unbalanced load conditions. The flexibility of the derived algorithm is such that facilities for online tap changing, the representation of saliency and induction motor loads may be incorporated.
10
ACKNOWLEDGMENTS

By way of example, the effects of opening one circuit of a double-circuit line at one end only are indicated for a 500 kV system comprising a number of line sections of long length with transpositions. Although the resulting unbalances are not particularly severe in this case, there are clearly other situations for which the 3-phase-load-flow program would prove invaluable as a means of investigating the possibility of large unbalance. In addition, the program could be used as a basis for studying unbalanced fault conditions. The program may be used to investigate the effects of unbalances in, say, machines and transformers. Furthermore, the influence of shunt and series compensation in both balanced and unbalanced form could be established together

The authors are grateful for the facilities provided by Prof.Wedepohl in the power systems laboratory of the University of Manchester Institute of Science & Technology, England, and to the University of Manchester Regional Computer Centre for running their programs. Mr.Shlash also expresses his gratitude to the National Electricity Administration of Iraq for support during his leave of absence. 11 REFERENCES 1 HESSE, M. H.: 'Circulating currents in paralleled untransposed lines. Inumerical evaluation', IEEE Trans., 1966, PAS-35, pp. 802-811

TABLE 4 BRANCH FLOWS AND PERCENTAGES OF NEGATIVE TO POSITIVE SEQUENCE CURRENTS FOR ABNORMAL LINE-END OPEN CONDITIONS

Sending end Sending busbar Receiving busbar Phases Active

MW

Receiving end In % 2-71 Active MW 270-44 280-78 26693 265-44 274-40 266-21 263-05 256-33 258-40 253-80 255-55 254-45 253-80 255-55 254-45 -0-80 0-95 -0-15 0-80 -0-95 0-15 Reactive
MVA

Reactive
MVA

In % 2-71

R Y B R Y B R Y B R Y B R Y B R Y B R Y B

275-55 27722 26539

270-44 280-78 266-93 265-44 274-40 266-21 263-05 256-33 258-40 253-80 255-55 254-45 -0-80 0-95 -0-15 1-08 -2-02 0-94

-52-39 -40-07 -41-73 -85-99 -92-70 -73-98 -111-75 -13080 -105-68 -129-05 -127-11 -136-86 -4-00 -6-76 -2-86 -104-31 -107-45 -103-17 114-11 112'76 111-21

-85-99 -92-70 -73-98 76-00 62-83 78-05 31-27 2768 22-35 -4-00 -6-76 -2-86 -20-11 -2325 -18-97 -107-54 -110-91 -106-32 107-54 110-91 106-32

2-71

1-70

2-35

0-55

0-75

0-70

0-70

0-70

1-64

1-64

1-64

1-64

TABLE 5 BUSBAR VOLTAGES AND PERCENTAGES OF VOLTAGE UNBALANCE FOR ABNORMAL LINE-END OPEN CONDITIONS R phase Busbar Magnitude p.u. 1 2 3 4
CJl

Y phase Angle degrees Magnitude p.u. 0-994 0-986 0-984 0-880 0912 0-916 0-945 1-001 Angle degrees -010 -38-18 -46-98 -71*63 -82-07 -85-75 -85-77 -55-74

B phase Magnitude p.u. 0-995 0-974 0-958 0-892 0-919 0-921 0-950 1-003 Angle degrees 0-26 -37-66 -4597 -73-00 -82-17 -85-78 -85-78 -55-87
V %

1-000 0-985 0-969 0-895 0-917 0-919 0-948 1-000

o-oo
-37-52 -45-76 -72-19 -82-32 -85-94 -85-93 -55-87

0-62 1-23 2-63 1-75 0-74 0-63 0-47 0-27

637

6 7 8

PROC. IEE, Vol. 121, No. 7, JULY 1974

SMITH, D. C, and STOKES, E. C: 'The effect of out of balance of transmission line impedance on the design of a 500 kV transmission network'. Proceedings of the 2nd power-systems computations conference, Stockholm, 1966, paper 4.15 EL-ABIAD, A. H.,. and TARSI, D. C: 'Load-flow solution of untransposed EHV networks'. Proceedings of the power-industry computer-applications conference, Pittsburgh, 1967 BRITTON, J. P.: 'Improved load-flow performance through a more general equation form'. IEEE Winter power meeting, New York, 1970, paper 70 TP 25-PWR DOMMEL, H. W., TINNEY, W. F., and POWELL, W. L.: 'Further developments in Newton-Raphson's method for power system applications'. IEEE Winter power meeting New York, 1970, paper 70 TP 25-PWR; paper 70 CP 161-PWR STOTT, B.: 'Effective starting process for NewtonRaphson load flow', Proc.IEE, 1971, 118, (9), pp. 983-987 STAGG, G. W., and EL-ABIAD, A. H.: 'Computer methods in power system analysis' (McGraw-Hill, 1968) LAUGHTON, M. A.: 'Analysis of unbalanced polyphase networks by the method of phase co-ordinates', Proc. IEE, 1968, 115, (8), pp. 1163-1172 ZOLLENKOPF: 'Bifactorisation'. Conference on large sets of sparse linear equations, Oxford, 1970 10 WEDEPOHL, L. M., and WASLEY, R. G.: 'Solution of travelling wave problems in multiconductor lines using

eigenvalue techniques'. Proceedings of the 3rd powersystems computation conference, Rome, 1969 APPENDIX For the purposes of the present paper the generator impedance matrix is simply derived from the sequence impedance matrix
12

Zo =

where Zo, Zx and Z 2 represent the zero, positive and negative sequence impedances, respectively. Thus, for the usual transformation matrix relating phase and sequence quantities, i.e. "l 1 1 1 a2 a l" a a2

it can be shown that the phase impedance matrix is given by

+ Z, + Z,

a2Z aZ,

Z o + a.2Z1 + aZ

ZG = TZ s T-i =

aZ

638

PROC. IEE, Vol. 121, No. 7, JULY 1974